UNet++ is a new architecture for medical image segmentation that improves upon the U-Net and wide U-Net. It uses nested, dense skip pathways to reduce the semantic gap between encoder and decoder feature maps, allowing the optimizer to handle an easier learning task. The architecture also incorporates deep supervision, which enables model pruning and improves segmentation accuracy. UNet++ was evaluated on four medical image segmentation tasks: nodule segmentation in low-dose CT scans, nuclei segmentation in microscopy images, liver segmentation in abdominal CT scans, and polyp segmentation in colonoscopy videos. The results showed that UNet++ with deep supervision achieved an average IoU gain of 3.9 and 3.4 points over U-Net and wide U-Net, respectively. The architecture also allows for pruning, which reduces inference time while maintaining segmentation accuracy. The proposed architecture is more effective than traditional skip connections in U-Net, which directly fast-forward high-resolution feature maps from the encoder to the decoder, resulting in the fusion of semantically dissimilar feature maps. UNet++ is a deeply supervised encoder-decoder network with re-designed skip pathways and deep supervision, which enables more accurate segmentation, particularly for lesions that appear at multiple scales. The architecture was tested on four medical imaging datasets and showed significant improvements in segmentation accuracy. The results demonstrate that UNet++ with deep supervision achieves better performance than U-Net and wide U-Net.UNet++ is a new architecture for medical image segmentation that improves upon the U-Net and wide U-Net. It uses nested, dense skip pathways to reduce the semantic gap between encoder and decoder feature maps, allowing the optimizer to handle an easier learning task. The architecture also incorporates deep supervision, which enables model pruning and improves segmentation accuracy. UNet++ was evaluated on four medical image segmentation tasks: nodule segmentation in low-dose CT scans, nuclei segmentation in microscopy images, liver segmentation in abdominal CT scans, and polyp segmentation in colonoscopy videos. The results showed that UNet++ with deep supervision achieved an average IoU gain of 3.9 and 3.4 points over U-Net and wide U-Net, respectively. The architecture also allows for pruning, which reduces inference time while maintaining segmentation accuracy. The proposed architecture is more effective than traditional skip connections in U-Net, which directly fast-forward high-resolution feature maps from the encoder to the decoder, resulting in the fusion of semantically dissimilar feature maps. UNet++ is a deeply supervised encoder-decoder network with re-designed skip pathways and deep supervision, which enables more accurate segmentation, particularly for lesions that appear at multiple scales. The architecture was tested on four medical imaging datasets and showed significant improvements in segmentation accuracy. The results demonstrate that UNet++ with deep supervision achieves better performance than U-Net and wide U-Net.